Summability methods and moment sequences. II. (Q1563231): Difference between revisions

The author develops what are now known as generalized Hausdorff matrices, as a solution to an interpolation problem. Let \(\{\lambda_n\}\) denote anonnegative sequence with \(0 = \lambda_0 < \lambda_1 < \cdots < \lambda_n< \cdots\) such that \(\sum_{n = 1}^{\infty} \nolimits 1/\lambda_n =\infty\). The entries of a generalized Hausdorff matrix are \(\lambda_{k+1}\cdots\lambda_n\Big(\frac{u^{t_k}\ldots,u^{t_n}}{\lambda_k\ldots\lambda_n}\Big)\), where the quantity in parentheses denotes the divided difference. For example, \(\Big(\frac{u^{t_{n-1}},u^{t_n}}{\lambda_{n-1}\lambda_n}\Big) = (u_{n-1} - u_n)/(\lambda_n -\lambda_{n_1})\). If the matrix preserves convergence, then \(\mu_n =\int_0^1u^{\lambda_n}d\chi(u)\), where \(\chi\) is a normalized function of bounded variation over \([0, 1]\).